Lab 7
Maximilian Halperin
Math 241, Week 10
# Put all necessary libraries here
library(tidyverse)
library(reprex)
library(dplyr)
library(infer)
library(moderndive)
library("styler")
library(lubridate)
library(ggthemes)
library(svglite)
library(DT)
library(shiny)
library(tidycensus)
library(ggmap)
library(leaflet)
library(tidycensus)
library(gganimate)
library(pdxTrees)
library(rvest)
library(httr)Due: Thursday, April 8th at 8:30am
Goals of this lab
Note: You should upload the final version to your Math 241 GitHub repo that I created for you, not Gradescope. We will grade this lab directly from your repo.
Goals of this lab
- Practice creating static and interactive choropleth maps.
- Practice creating static maps with
ggmapand interactive maps withleaflet. - Practice incorporating interactivity into a graph with
gganimate.
Problem 1: Mapping PDX Crashes
For this problem we will return to the SE Portland 2018 car crash dataset.
pdx_crash_2018 <- read_csv("/home/courses/math241s21/Data/pdx_crash_2018_page1.csv")- Grab the code from Lab 2, Problem 4.a to create a scatterplot of longitude and latitude (
LONGTD_DD,LAT_DD). Paste that code here to recreate that graph.
ggplot(data = pdx_crash_2018,
mapping = aes(
x = LONGTD_DD,
y = LAT_DD,
color = "#4C4CFD")) +
geom_point(alpha = 0.3)
- Now create a (static) raster map with the crashes mapped as points on top.
box <- c(bottom = 45.45, left = -122.7, top = 45.54, right = -122.45)
reed <- get_stamenmap(box, maptype = "toner", zoom = 12)
reed %>%
ggmap() +
geom_point(aes(x = LONGTD_DD, y = LAT_DD), data = pdx_crash_2018,
color = "red", size = 3)
- Now create an interactive map of the crashes (but still only map the location of the crashes at this point).
#leaflet(options = leafletOptions(minZoom = 11, maxZoom = 13.5)) %>%
leaflet() %>%
setView(lng = -122.58, lat = 45.495, zoom = 11) %>%
addTiles() %>%
addCircleMarkers(lng = ~LONGTD_DD, lat = ~LAT_DD,
data = pdx_crash_2018, radius = 1,
opacity = 0.2)- Create a factor variable,
day_timewith categories:
pdx_time <- pdx_crash_2018 %>%
filter(CRASH_HR_NO != 99) %>%
mutate(CRASH_HR_NO = as.numeric(CRASH_HR_NO),
day_time = case_when(CRASH_HR_NO > 5 & CRASH_HR_NO <= 12 ~ "Morning",
CRASH_HR_NO > 12 & CRASH_HR_NO <= 16 ~ "Afternoon",
CRASH_HR_NO > 16 & CRASH_HR_NO <= 20 ~ "Evening",
CRASH_HR_NO > 20 & CRASH_HR_NO <= 23 ~ "Night",
CRASH_HR_NO <= 5 ~ "Night"),
day_time = as.factor(day_time),
CRASH_TIME = fct_relevel(day_time, "Night", "Evening", "Afternoon", "Morning")
)
factpal <- colorFactor("PuOr", pdx_time$day_time)
pdx_time %>%
leaflet() %>%
setView(lng = -122.58, lat = 45.495, zoom = 11) %>%
addTiles() %>%
addCircleMarkers(lng = ~LONGTD_DD, lat = ~LAT_DD,
data = pdx_time,
color = ~factpal(day_time),
stroke = FALSE, fillOpacity = 0.5,
radius = 3.5) %>%
addLegend("bottomright", pal = factpal,
values = ~day_time, title = "Crash",
opacity = 1)- morning: crashes between 5am and noon
- afternoon: crashes after noon but before or at 4pm
- evening: crashes after 4pm but before or at 8pm
- night: crashes after 8pm but before or at 5am
Add this variable to your interactive map using color. Make sure to include a legend and be mindful of your color palette choice. How do crash locations vary by parts of the day?
- Now add a pop-up to you interactive map that provides additional information (beyond part of the day) about the crash.
pdx_crash_2018 <- pdx_crash_2018 %>%
mutate(CRASH_DT = str_remove(CRASH_DT, "00:00:00"))
date_crash <- paste("Date of Crash:", pdx_crash_2018$CRASH_DT)
pdx_time %>%
leaflet() %>%
setView(lng = -122.55, lat = 45.5, zoom = 12) %>%
addTiles() %>%
addCircleMarkers(lng = ~LONGTD_DD, lat = ~LAT_DD,
data = pdx_time, color = ~factor(day_time),
stroke = FALSE, fillOpacity = 0.5,
radius = 3.5, popup = date_crash)- Create a leaflet graph where the user can toggle between displaying the different crash severities. Draw some conclusions about differences in location for the severity types.
sev1 <- pdx_crash_2018 %>%
mutate(severity = case_when(
CRASH_SVRTY_CD == 2 ~ "fatal")) %>%
filter(severity == "fatal")
sev2 <- pdx_crash_2018 %>%
mutate(severity = case_when(
CRASH_SVRTY_CD == 4 ~ "injury")) %>%
filter(severity == "injury")
sev3 <- pdx_crash_2018 %>%
mutate(severity = case_when(
CRASH_SVRTY_CD == 5 ~ "property damage")) %>%
filter(severity == "property damage")
leaflet() %>%
setView(lng = -122.55, lat = 45.5, zoom = 12) %>%
addTiles() %>%
addCircleMarkers(lng = ~LONGTD_DD, lat = ~LAT_DD,
data = sev1, color = "red",
stroke = FALSE, fillOpacity = 0.5,
radius = 3.5, popup = date_crash, group = "fatal") %>%
addCircleMarkers(lng = ~LONGTD_DD, lat = ~LAT_DD,
data = sev2, color = "blue",
stroke = FALSE, fillOpacity = 0.5,
radius = 3.5, popup = date_crash, group = "injury") %>%
addCircleMarkers(lng = ~LONGTD_DD, lat = ~LAT_DD,
data = sev3, color = "green",
stroke = FALSE, fillOpacity = 0.5,
radius = 3.5, popup = date_crash, group = "property damage") %>%
addLayersControl(overlayGroups = c("fatal", "injury","property damage"))Looks like fatal crashes occur near the highway and freeway while property damage incidents are concentrated downtown and the major roads. g. Let’s go back to the static map and this time change our geom. Use geom_density2d() instead of geom_point(). Interpret what this map tells us about car crashes in the SE and compare the story to the map using geom_point().
box <- c(bottom = 45.45, left = -122.7, top = 45.54, right = -122.45)
reed <- get_stamenmap(box, maptype = "toner", zoom = 12)
reed %>%
ggmap() +
geom_density2d(aes(x = LONGTD_DD, y = LAT_DD), data = pdx_crash_2018,
color = "red")
Compared to the static map, the number of rings in a particular location indicates more frequent accidents in the location. We can see that near highways and near downtown there are more accidents compared to other areas.
- Facet the plot in g. by
day_time. Does the distribution on accidents seem to vary much by part of day?
reed %>%
ggmap() +
geom_density2d(data = pdx_time, aes(x = LONGTD_DD, y = LAT_DD),
inherit.aes = FALSE,
color = "red") +
facet_grid(day_time ~ .) +
theme_minimal()
Areas with high accident distribution do not change. ### Problem 2: Choropleth Maps
For this problem, I want you to practice creating choropleth maps. Let’s grab some data using tidycensus. Remember that you will have to set up an API key.
api_key <- "6413ba5315708be54be79c9d03ead7b0a4b29d3a"
library(tidycensus)- Let’s grab data on the median gross rent (
B25064_001) from the American Community Survey for Multnomah county, Oregon. I want you to do data pulls at three geography resolutions:county subdivision,tract, andblock group.
rent_county <- get_acs(geography = "county subdivision",
variables = "B25064_001",
county = "multnomah",
state = "Oregon",
geometry = TRUE,
key = api_key,
cache_table = TRUE)##
|
| | 0%
|
|== | 2%
|
|== | 3%
|
|=== | 4%
|
|=== | 5%
|
|==== | 6%
|
|===== | 7%
|
|====== | 8%
|
|====== | 9%
|
|======== | 11%
|
|======== | 12%
|
|========== | 14%
|
|=========== | 15%
|
|=========== | 16%
|
|============ | 17%
|
|============= | 18%
|
|============== | 20%
|
|================ | 23%
|
|=================== | 28%
|
|====================== | 32%
|
|======================== | 34%
|
|=========================== | 39%
|
|================================================== | 71%
|
|===================================================== | 76%
|
|=========================================================== | 85%
|
|======================================================================| 100%tract_county <- get_acs(geography = "tract",
variables = "B25064_001",
county = "multnomah",
state = "Oregon",
geometry = TRUE,
key = api_key,
cache_table = TRUE)##
|
| | 0%
|
|======================================================================| 100%block_county <- get_acs(geography = "block group",
variables = "B25064_001",
county = "multnomah",
state = "Oregon",
geometry = TRUE,
key = api_key,
cache_table = TRUE)##
|
| | 0%
|
|============================================================= | 86%
|
|======================================================================| 100%- Create three choropleth maps of gross rent, one for each geography resolution. What information can we glean from these maps? Also, which resolution seems most useful for this variable? Justify your answer.
ggplot(data = rent_county, mapping = aes(geometry = geometry, fill = estimate)) +
geom_sf() +
coord_sf() +
scale_fill_viridis_c(direction = -1) +
theme_void()
ggplot(data = tract_county, mapping = aes(geometry = geometry, fill = estimate)) +
geom_sf() +
coord_sf() +
scale_fill_viridis_c(direction = -1) +
theme_void()
ggplot(data = block_county, mapping = aes(geometry = geometry, fill = estimate)) +
geom_sf() +
coord_sf() +
scale_fill_viridis_c(direction = -1) +
theme_void()
Tract seems to be the vest resolution. block_county is missing a lot data on gross rent and the divisions are small. The county subdivision only has five gross rent divisions which does not accurately represent Multnomah county. tract_county is the best because it has mostly complete data and the subdivisions do are large enough for the viewer to identify the area.
- Make one of your maps interactive.
pal4 <- colorFactor(palette = "viridis", domain = rent_county$estimate)
rent_county %>%
sf::st_transform(crs = "+init=epsg:4326") %>%
leaflet() %>%
addTiles() %>%
addPolygons(popup = ~estimate, fillColor = ~pal4(estimate),
stroke = FALSE, fillOpacity = 0.9) %>%
addLegend("bottomright", pal = pal4,
values = ~estimate, title = "Median Income",
opacity = 1)Problem 3: Take a Static Plot and Animate It!
Let’s take a static plot we made earlier in the semester and add some animation.
- Go through your previous labs or mini-projects and find a plot you think might be better with animation. Grab the code and recreate the graph here.
.
grad_rate <- "https://www.reed.edu/ir/gradrateshist.html"
rate_table <- grad_rate %>%
read_html() %>%
html_nodes(css = "table")
grad_rate_table <- html_table(rate_table[[1]], fill = TRUE)
colnames(grad_rate_table) <- c("Year", "Cohort_size", "gradfour", "gradfive", "gradsix")
grad_rate_table <- grad_rate_table %>%
slice(-1)
gradratetable <- grad_rate_table %>%
mutate(Grad_year = as.numeric(Year),
cohort = parse_number(Cohort_size),
four = parse_number(gradfour),
five = parse_number(gradfive),
six = parse_number(gradsix)
) %>%
dplyr::select(Grad_year, cohort, four, five, six)
grad_rate_table_final <- pivot_longer(gradratetable, cols = c(four, five, six),
names_to = "Years to Graduate",
values_to = "Graduation Rate") %>%
select(Grad_year, cohort, `Years to Graduate`, `Graduation Rate`)
static <- ggplot(grad_rate_table_final, aes(fill = `Years to Graduate`, y = `Graduation Rate`, x = Grad_year)) +
geom_bar(stat = "identity") +
facet_grid(. ~ `Years to Graduate`) +
theme_bw() +
labs(x = "Graduation Year")- Now add animation.
animation <- static +
transition_manual(Grad_year, cumulative = TRUE)
#transition_reveal(along = Grad_year)
animate(animation, fps = 10, end_pause = 20)
- In what ways did the animation improve the plot? In what ways did the animation worsen the plot?
It really shows the growth over time for each group and learn which groups stay relatively consistent over time(people who graduate in six years). Depending the preferences of the viewer, animation made the plot more difficult compare the graduation rates as there is no static point in the animation.
Problem 4: Your Turn!
- Using the
pdxTreesdataset, create twoleafletmaps. I want you to get creative and really dig into the functionalities ofleaflet. Consider
- Focusing on a specific park or set of parks.
- Potentially using special icons.
- Including labels and/or pop-ups.
- The best tiling for your purpose.
- Zoom or view constraints
parks <- get_pdxTrees_parks()
parks <- parks %>%
filter(Common_Name == c("Douglas-Fir", "Bigleaf Maple", "American Sycamore", "American Elm", "Austrian Black Pine"))
pal3 <- colorFactor(palette = "viridis", domain = parks$Common_Name)
parks %>%
leaflet() %>%
addTiles() %>%
addCircleMarkers(lng = ~ Longitude, lat = ~Latitude,
data = parks,
stroke = FALSE,
fillOpacity = 0.6,
color = ~pal3(Common_Name)) %>%
addLegend("bottomright", pal = pal3,
values = ~Common_Name, title = "Tree",
opacity = 1) %>%
setView(lng = -122.629383, lat = 45.481245, zoom = 11)State the key takeaways that we can learn from your maps. The graph can show the viewer how diverse the species of trees are in parks. It also can tell us where the trees are located.
- Now using the
pdxTreesdataset, create an animated graph. Again, think carefully about the various animation features. In particular, consider
- How you want to transition from frame to frame.
- How the data should enter and exit the plot.
- The speeds of various aspects of the animation.
- Adding frame information to the title and/or subtitle.
- Whether or not the view should change as the animation progresses.
pdxpal <- colorFactor("Tree Common Name:", parks$Common_Name)
name <- paste("Tree Common Name:", parks$Common_Name)
something <- parks %>%
mutate(Condition = fct_relevel(Condition, "Good", "Fair", "Poor")) %>%
ggplot(data = parks, mapping = aes(x = Longitude, y = Latitude, color = Condition)) +
geom_point() +
transition_manual(Inventory_Date, cumulative = TRUE)
animate(something, fps = 5, end_pause = 20)
State the key takeaways that we can learn from your animated graph. Also address whether or not you think the animation helps or hinders the delivery of these key takeaways. I was trying to show tree condition over time but wasn’t able to show that. I think it hinders the key takeaway due to the map not being portrayed.